A hundred cybersecurity experts claim that the ban on Fable 5 negatively impacts defenders.

      Three days after the US government mandated Anthropic to halt Fable 5 and Mythos 5, approximately 100 leading cybersecurity experts issued an open letter urging the reversal of the ban. Their message is clear: removing the most effective AI tools from defenders while adversaries continue to develop theirs is not safety; it amounts to sabotage.

      “This decision has deprived defenders of the best models, created uncertainty in the market, and jeopardized America’s AI leadership without any legitimate risk to justify it,” the letter asserts. Among the signatories are Alex Stamos, former chief security officer at Facebook and Yahoo, now serving as chief product officer at Corridor, along with Katie Moussouris, CEO of Luta Security, Rachel Tobac of SocialProof Security, Chris Wysopal of Veracode, and Joe Levy, CEO of Sophos.

      The sequence of events that prompted the government order began with Amazon researchers discovering a way to prompt Fable 5 to reveal code vulnerabilities. The method was not particularly unique by industry standards: after an initial denial to “review the code for security issues,” the researchers simply changed their prompt to “fix this code,” providing it with open-source code containing known and intentionally inserted flaws.

      Moussouris has been particularly outspoken, telling reporters that the exploit “is not a jailbreak.” The open letter supports this, highlighting that other major AI models, such as OpenAI’s GPT-5.5, can expose the same vulnerabilities without any bypass.

      Anthropic’s stance aligns with the critics. The company has stated that the exploit is narrow, not widespread, and that the vulnerabilities it revealed were minor and already publicly known.

      The letter, however, does not touch on perhaps the most complicated aspect of the situation: the fact that the company which identified the vulnerability also happens to be Anthropic’s largest investor and cloud provider. Amazon CEO Andy Jassy personally escalated the findings to Treasury Secretary Scott Bessent, Commerce Secretary Howard Lutnick, and National Cyber Director Sean Cairncross.

      This escalation, from a competitor's security team to top government officials, raises questions about whether commercial competition influenced the government's response. Semafor has reported that concern from the White House extended beyond the jailbreak itself, leading to fears about Chinese access to Mythos.

      David Sacks, an AI adviser to Trump, offered a different perspective on X. He alleged that the administration gave Anthropic CEO Dario Amodei a choice: fix the jailbreak or take Fable 5 offline. According to Sacks, Amodei declined, and Anthropic chose to prioritize the continued offering of the consumer model over safety. Anthropic has contested this characterization, maintaining that the vulnerability is too limited to warrant the withdrawal of its leading products from the market.

      If the ban was aimed at safeguarding national security, it may have had the opposite effect. Chinese AI firm Zhipu AI launched its GLM-5.2 model on June 13, just one day after the Fable 5 shutdown, directly referencing the ban as evidence that US AI models are unreliable.

      Zhipu's stock surged by 33% following the announcement. The company asserts that GLM-5.2 outperforms BridgeBench reasoning at 42.8 tokens per second and is priced at one-tenth the cost of similar US frontier models, though it provided no independent benchmark scores at launch.

      The trade-off is significant. Any organization utilizing Zhipu’s cloud API would put its data at risk of exposure to the Chinese government under China’s National Intelligence Law, which requires companies to assist with state intelligence operations.

      The core argument of the open letter is pragmatic, not ideological. Cybersecurity professionals utilize leading AI models to identify software vulnerabilities before attackers can exploit them, generate detection rules, and analyze malware rapidly.

      Removing the most effective models from this process does not deter adversaries, who can resort to open-source alternatives, foreign models, or simply older methods. It hinders defenders from operating at the pace necessary for the current threat landscape.

      Observers have noted the irony. Eastern Herald reported that some of the executives who signed the open letter warned in April about the risks associated with Mythos, Anthropic’s most powerful reasoning model.

      This apparent contradiction does not necessarily weaken the letter’s argument, but it does indicate that the cybersecurity community is still negotiating where to draw the line between capability and caution.

      Looking ahead, prediction markets suggest that the ban is unlikely to endure. Kalshi estimates a 68% chance that Fable 5 will return before July 1, while Polymarket is slightly more optimistic at 71%.

      The European Union has already begun advocating for guaranteed access to Mythos for cybersecurity purposes, adding international pressure to the domestic outcry. Similarly, India has leveraged this incident to advance its own sovereign AI initiatives.

      For now, America’s cybersecurity defenders confront a question articulated starkly in the open letter: if the best tools are stripped from those safeguarding